**** BEGIN LOGGING AT Thu Jan 12 03:00:21 2006
Jan 12 10:25:57 <lennert> the ixp23xx built-in 10/100 ethernet also seems to require NPE/access lib support
Jan 12 10:26:05 <lennert> so i might end up having a go at that
Jan 12 10:57:45 <ka6sox-laptop> lennert: any word from Fabrics?
Jan 12 10:58:43 <lennert> ka6sox-laptop: yeah -- they might send me a loaner board if i also do some kernel work for them
Jan 12 10:58:56 <lennert> ka6sox-laptop: i'll have to do kernel work anyway since the board doesn't run 2.6 at all yet
Jan 12 10:59:44 <lennert> ka6sox-laptop: i'll hear back from them somewhere These Days i hope
Jan 12 11:01:48 <ka6sox-laptop> COOL!
Jan 12 11:24:15 <prpplague> beewoolie-afk: hey hey
Jan 12 11:24:41 <beewoolie-afk> prpplague: yo.
Jan 12 11:24:56 <beewoolie-afk> prpplague: I've been refining the basics of my JTAG test code.
Jan 12 11:25:04 <beewoolie-afk> Today, we see if I can access the debug unit.
Jan 12 11:29:24 <prpplague> beewoolie-afk: hehe cool
Jan 12 11:29:39 <prpplague> i've not had time to tinker alot with the debug stuff
Jan 12 11:36:21 <prpplague> beewoolie: hehe i was hopin to be finished with this legacy usb crap yesterday, but i looks like i wont be back dev'n till monday
Jan 12 11:36:36 <prpplague> beewoolie: should be able to get caught up on the s3c2410 patch and send you some additional code
Jan 12 11:37:02 <beewoolie> That would be nice.
Jan 12 11:37:32 <prpplague> beewoolie: i noticed that the changelog for both .11 and .12 have references to the lnode80
Jan 12 11:37:48 <prpplague> beewoolie: did something change between .11 and .12 ?
Jan 12 11:37:54 <beewoolie> Let me see...
Jan 12 11:39:01 <beewoolie> I had to recode the timer.  I modified the lnode80 version as well as the others.  I wanted to use proper __asm syntax so that register allocation was handled by the compiler.   Moreover, I change the constraints to be more correct.
Jan 12 11:39:12 <beewoolie> The code shouldn't have changed.
Jan 12 11:39:20 <beewoolie> That is the output code shouldn't have changed.
Jan 12 11:39:58 <prpplague> ahh ok
Jan 12 11:40:04 <prpplague> beewoolie: just wanted to double check
Jan 12 16:13:35 <ulf_k> hi
Jan 12 16:13:53 <ulf_k> does openjtag support ixp42x ? 
Jan 12 16:21:18 <vmaster> uhm... there isn't "openjtag" as a product that could support anything
Jan 12 18:01:22 <beewoolie> ulf_k_: I wonder if you are asking about the openwince jtag software
Jan 12 18:01:41 <beewoolie> vmaster: Are you on?
Jan 12 18:02:14 <ulf_k_> hi beewoolie
Jan 12 18:02:25 <beewoolie> hi
Jan 12 18:06:04 <ulf_k_> well later on i realized that openwince and openjtag are not the same
Jan 12 18:08:03 <lennert> openjtag isn't much, so far
Jan 12 18:08:19 <lennert> nothing but untold megabytes of irc logs :)
Jan 12 18:08:33 <lennert> and a bit of code here and there
Jan 12 18:16:14 <beewoolie> lennert: Hi lennert
Jan 12 18:16:31 <beewoolie> lennert: I'm doing a timing test on the code as implemented in openwince
Jan 12 18:16:42 <beewoolie> It is interesting to note that the usleep function really blows.
Jan 12 18:17:28 <lennert> hi beewoolie
Jan 12 18:17:29 <lennert> hehe
Jan 12 18:17:39 <lennert> pmc3386 driver from radisys did something like
Jan 12 18:17:51 <lennert> os_linux_enp_udelay(int MicroSeconds)
Jan 12 18:17:53 <lennert> {
Jan 12 18:17:56 <lennert> int i;
Jan 12 18:18:03 <lennert> for (i=0;i<MicroSeconds*50000;i++)
Jan 12 18:18:04 <lennert> ;
Jan 12 18:18:04 <lennert> }
Jan 12 18:18:12 <beewoolie> Um, ....yeah.
Jan 12 18:18:24 <beewoolie> In a word...poop.
Jan 12 18:18:32 <lennert> hell yeah
Jan 12 18:19:01 <beewoolie> It turns out that without a delay, we get about 268KHz.
Jan 12 18:19:14 <beewoolie> Moreover, I am using a very different design from openwince.
Jan 12 18:19:15 <lennert> parallel cble?
Jan 12 18:19:26 <beewoolie> My inner loop is much faster.
Jan 12 18:19:39 <beewoolie> At the moment, I'm testing the Xilinx III.
Jan 12 18:19:48 <beewoolie> It makes it possible for me to verify my code.
Jan 12 18:19:55 <lennert> right
Jan 12 18:20:07 <lennert> i could download the spartan3 via jtag at 700khz or so
Jan 12 18:20:14 <lennert> but i think that pc had a fast parallel port
Jan 12 18:20:25 <lennert> http://svn.wantstofly.org/vhdl/utils/  (very incomplete)
Jan 12 18:20:51 <beewoolie> Well, keep in mind that I'm looking at the clock signal.
Jan 12 18:21:03 <beewoolie> This doesn't say anything about the data throughput.
Jan 12 18:21:31 <lennert> i'm still not 100% sure what you're doing.. are you running openwince and looking at TCK with a scope?
Jan 12 18:21:41 <beewoolie> Moreover, I'm using the PPDEV interface which, I suspect, isn't as fast as the direct port access interface.
Jan 12 18:21:50 <beewoolie> Actually, no.
Jan 12 18:22:02 <beewoolie> I'm working on code to manipulate the JTAG scan chains.
Jan 12 18:22:04 <lennert> oh, yeah -- i'm just bitbanging the parallel port i/o ports from userspace (ugh ugh)
Jan 12 18:22:12 <beewoolie> I can read registers and do lots of interesting things already.
Jan 12 18:22:23 <beewoolie> I decided that I wanted to see what the throughput limit was.
Jan 12 18:22:41 <lennert> so time how long an MMIO cycle takes?
Jan 12 18:22:42 <beewoolie> Well, the goal is much grander.
Jan 12 18:22:56 <beewoolie> MMIO?  I don't know what that means in this context.
Jan 12 18:23:07 <lennert> sorry, not mmio
Jan 12 18:23:14 <lennert> time how long it takes to do a parallel port write, i mena
Jan 12 18:23:24 <lennert> that gives you the throughput limit
Jan 12 18:23:30 <lennert> or am i still not understanding you correctly :-/
Jan 12 18:24:42 <beewoolie> Well, the inner loop of this test is write clock high, write clock low, repeat.
Jan 12 18:24:50 <beewoolie> I'm getting about 268KHz.
Jan 12 18:25:03 <lennert> right, that's what i did too
Jan 12 18:25:04 <beewoolie> I'm going to see what happens if I use the port directly without ppdev.
Jan 12 18:25:09 <lennert> i got 700khz
Jan 12 18:25:15 <beewoolie> Perhaps your machine is faster.
Jan 12 18:25:23 <beewoolie> Are you using ppdev?
Jan 12 18:25:29 <lennert> try http://svn.wantstofly.org/vhdl/utils/test_speed.c  ;-)
Jan 12 18:25:41 <lennert> no, direct port i/o, which makes it much faster as it avoids the kernel round-trip
Jan 12 18:25:48 <lennert> but much less portable so not a good solution
Jan 12 18:26:03 <beewoolie> Um, that code doesn't do anything.
Jan 12 18:26:09 <beewoolie> Whatsa jtag_flush?
Jan 12 18:26:16 <lennert> jtag.c should be called xilinx_parallel3_cable_specific_hacks.c
Jan 12 18:26:20 <lennert> see jtag.c in that directory
Jan 12 18:26:29 <lennert> i'm ashamed to even show you this stuff
Jan 12 18:26:49 <beewoolie> OK.  That's using the port directly.
Jan 12 18:27:07 <beewoolie> I'm using the user-mode interface at the moment.  The code runs without root privileges.
Jan 12 18:27:20 <lennert> which is a big plus
Jan 12 18:27:32 <beewoolie> Except that what we really care about is speed.
Jan 12 18:27:35 <lennert> i'm just saying, if you want to benchmark pure port accesses you can use that
Jan 12 18:27:43 <lennert> instead of having to hack it up yourself
Jan 12 18:27:46 <lennert> well, true
Jan 12 18:27:47 <beewoolie> I think I might look into using root to gain access to the port and then dropping priviledges.
Jan 12 18:27:55 <lennert> you need root then though
Jan 12 18:27:59 <beewoolie> Ah, right.
Jan 12 18:28:03 <lennert> you could use root if you have it and ppdev otherwise
Jan 12 18:28:13 <beewoolie> Or, we can make the exec file be setuid.
Jan 12 18:28:51 <beewoolie> Let me see what your code does...
Jan 12 18:29:28 * beewoolie loves wget.
Jan 12 18:29:49 <beewoolie> Is this a subversion repo being exposed?
Jan 12 18:29:49 * lennert loves subversion-web
Jan 12 18:29:52 <lennert> yes
Jan 12 18:29:58 <lennert> you can check it out with svn checkout http:// etc
Jan 12 18:30:02 <beewoolie> Ah,  Interesting.  I use subversion, but I've never exported anything.
Jan 12 18:30:09 <lennert> mod_dav_subversion iirc
Jan 12 18:30:16 <beewoolie> Hmm.  let me try that...
Jan 12 18:30:18 <lennert> plugs directly into apache
Jan 12 18:31:20 <beewoolie> elf@cerise ~...lennert/utils > sudo ./test_speed
Jan 12 18:31:20 <beewoolie> parallel jtag cable not installed
Jan 12 18:31:24 <beewoolie> Ooops.
Jan 12 18:31:30 <lennert> it's not in lpt1?
Jan 12 18:31:31 <beewoolie> I'll bet I know what it is.
Jan 12 18:32:07 <lennert> my digilent cable shorts some input and some output pin
Jan 12 18:32:08 <lennert> so it checks that
Jan 12 18:32:39 <beewoolie> It just exits.
Jan 12 18:32:51 <lennert> yeah
Jan 12 18:32:53 <lennert> bluntly
Jan 12 18:33:56 <beewoolie> Um, I don't get it.  Are you just timing the loop?
Jan 12 18:33:59 <lennert> yeah
Jan 12 18:34:08 <beewoolie> ah.  I've got a scope on it....
Jan 12 18:34:21 <lennert> you can put a "tck ^= 1;" in the inner loop if you want
Jan 12 18:34:31 <lennert> but whether you write tck as 0 0 0 0 or 0 1 0 1 won't make difference in speed
Jan 12 18:34:36 <lennert> i just use 'time' to time 16777216 writes
Jan 12 18:35:24 <beewoolie> OK.  It's running at 326KHz.  Faster, but not twice as fast.  Interesting.
Jan 12 18:35:42 <lennert> it depends a lot on your motherboard chipset
Jan 12 18:35:48 <beewoolie> Thanks.  That saved me some time.  I'll probably still support the direct access mode just for the fun of it.
Jan 12 18:35:54 <beewoolie> How is that?
Jan 12 18:35:55 <lennert> specifically, the 'distance' between your cpu and southbridge
Jan 12 18:36:04 <beewoolie> Isn't this just a matter of the speed of the CPU?
Jan 12 18:36:06 <lennert> i.e. how long it takes to do a write to a register in the southbridge
Jan 12 18:36:10 <lennert> nooo, not at all
Jan 12 18:36:18 <lennert> my pentium 200 is likely faster at this than your p4 3.6ghz
Jan 12 18:36:24 <beewoolie> Interesting.
Jan 12 18:36:34 <beewoolie> This is a notebook running at 1.6GHz.
Jan 12 18:36:41 <lennert> for ages, robert olsson fastest routing box was a supermicro 933mhz pentium iii
Jan 12 18:36:47 <beewoolie> It's not a p4.  It's a pentium M, I believe.
Jan 12 18:36:50 <lennert> why?  because the pci access latency is so damn low.
Jan 12 18:37:00 <lennert> it creamed the xeons in every test
Jan 12 18:37:11 <beewoolie> I have a supermicro box, too.  It was really great, but the athlon 64 single proc was still faster.
Jan 12 18:37:23 <lennert> in this test, the cpu is stalling thousands of cycles for each I/O port write.. it really is that slow.
Jan 12 18:37:23 <beewoolie> Ah.  Nice.  I like that company's machines.
Jan 12 18:37:35 <beewoolie> I've been waiting for them to release a dual-proc AMD machine.
Jan 12 18:37:40 <beewoolie> ...that I can afford.
Jan 12 18:37:59 <lennert> 1.6 GHz / 326 kHz = a whopping 4907 cycles per parallel port i/o write
Jan 12 18:38:47 <lennert> if i get one board i was hoping to get, i'll need a pci-express mobo to put it in
Jan 12 18:38:50 <beewoolie> Even giving instructions 10 cycles per execute, we are at 500 to 1.
Jan 12 18:39:02 <lennert> so i'll have to buy a new mobo, and might as well just buy a nice AMD box anyway
Jan 12 18:39:02 <beewoolie> Actually, 500 to 2 since we are writing twice.
Jan 12 18:39:30 <beewoolie> This also helps with understanding the real advantages of the USB dongle.
Jan 12 18:39:35 <lennert> an i/o port write will likely break the pipeline completely and put the cpu in microcode
Jan 12 18:39:50 <lennert> that could be a couple hundred cycles
Jan 12 18:39:51 <beewoolie> I don't see what you mean.
Jan 12 18:40:05 <beewoolie> Oh, you're saying that the IO itself is really really inefficient.
Jan 12 18:40:08 <lennert> then to get the write to the northbridge.. to the southbridge.. and wait for the ack
Jan 12 18:40:12 <lennert> yes
Jan 12 18:40:15 <lennert> the i/o itself is dog slow
Jan 12 18:40:26 <beewoolie> that's why memory mapped IO is so nice.
Jan 12 18:40:32 <lennert> most PC NICs use memory-mapped I/O (i.e. memory transactions instead of i/o transactions) because of that
Jan 12 18:40:35 <lennert> yeah :)
Jan 12 18:40:48 <beewoolie> The port thing has been stupid for a really long time.
Jan 12 18:40:51 <lennert> well, on most embedded boards pci i/o is done with memory transactions anyway.. but that's a whole nother sotry
Jan 12 18:41:00 <beewoolie> I can kinda see it when there is only 64K of address space.
Jan 12 18:41:06 <beewoolie> Right.
Jan 12 18:41:13 <beewoolie> This IO port stuff is legacy.
Jan 12 18:41:15 <lennert> yeah
Jan 12 18:41:23 <lennert> and pc makers don't bother optimising i/o accesses at all
Jan 12 18:41:44 <lennert> (not that pci memory latency is much better, but oh well...)
Jan 12 18:42:07 <lennert> latency has been steadily going down the tubes
Jan 12 18:42:18 <lennert> bandwidth keeps improving by leaps and bounds, but latency..
Jan 12 18:42:22 <lennert> latency of latency
Jan 12 18:42:28 <lennert> s/of/oh/
Jan 12 18:42:28 <purl> lennert meant: latency oh latency
Jan 12 18:43:21 <beewoolie> It's really killing us.
Jan 12 18:43:31 <beewoolie> Makes it even more important to get more cores.
Jan 12 18:44:20 <lennert> so that we can sacrifice half of the cores to stalling on i/o :)
Jan 12 18:44:34 <beewoolie> Right!
Jan 12 18:44:36 <lennert> and the other ones to our firewall software and virus and malware scanners
Jan 12 18:44:48 <lennert> and Word will still run slow
Jan 12 18:44:50 <beewoolie> More cores!  More cores!  Bring on the cores!!!!!
Jan 12 18:44:58 <lennert> haha
Jan 12 18:45:00 <beewoolie> CORES CORES CORES
Jan 12 18:45:04 <lennert> i'm all for more processing power :)
Jan 12 18:45:06 <beewoolie> SPAM SPAM SPAM SPAM....
Jan 12 18:45:21 <lennert> :))
Jan 12 18:45:23 <beewoolie> Wonderful cores!  Wonderful cores!
Jan 12 18:45:45 <lennert> get your cheap cores!  $100 a hundred!
Jan 12 18:45:49 <beewoolie> I saw an interesting link about a really small power supply...
Jan 12 18:46:26 <lennert> a power supply for each core? :)
Jan 12 18:46:44 <beewoolie> Looking...
Jan 12 18:46:50 <lennert> i'm very interesting in efficient hardware
Jan 12 18:46:56 <lennert> s/interesting/interested/
Jan 12 18:46:56 <purl> lennert meant: i'm very interested in efficient hardware
Jan 12 18:46:57 <beewoolie> http://www.bit-tech.net/news/2006/01/07/pico_psu/
Jan 12 18:47:11 <beewoolie> What I don't get is why this is interesting.
Jan 12 18:47:25 <lennert> before i got the epia i was thinking about making my internet gateway an arm box
Jan 12 18:47:47 <lennert> 120 watt.. damn, pretty small thing
Jan 12 18:47:55 <beewoolie> That's been on my agenda, too.  I got close, but the slug crashes.
Jan 12 18:48:10 <beewoolie> 10A at 12V is OK by me.
Jan 12 18:48:10 <lennert> and my slug has amnesia
Jan 12 18:48:23 <lennert> i was thinking of a simtec atx or mini-itx board
Jan 12 18:48:34 <lennert> but in the end the epia was cheaper, and doesn't draw much more power anyway
Jan 12 18:49:11 <beewoolie> I've been most interested in low power -> low noise.
Jan 12 18:49:17 <lennert> same
Jan 12 18:49:26 <lennert> my previous box was a dual xeon
Jan 12 18:49:37 <lennert> booting it was like firing up a jet
Jan 12 18:50:21 <beewoolie> My most common maintenance task has been replacing fans.
Jan 12 18:51:20 <lennert> i have an atx-size case with 4x200g disks and i just don't know how to keep it cool other than leaving the side panels off and pointing a big desk fan at it
Jan 12 18:51:56 <lennert> the epia board has a fan on the cpu, but since it's a 1.0ghz and i always run it at 500mhz, and since the psu fan _and_ the case fan blow/suck the exact area where the cpu, i can probably rip the cpu fan off
Jan 12 18:52:18 <lennert> (epia is a different box)
Jan 12 22:26:16 <beewoolie> Quiet day.
Jan 12 22:28:49 <ka6sox-laptop> ya...Hi beewoolie 
Jan 12 22:28:57 <beewoolie> ka6sox-laptop: Hey man.
Jan 12 22:29:14 <ka6sox-laptop> are you able to get all the parts for ep1220's interface?
Jan 12 22:29:18 <beewoolie> ka6sox-laptop: Did I tell you I repaired the JTAG you sent me?
Jan 12 22:29:26 <ka6sox-laptop> what was wrong with it?
Jan 12 22:29:29 <beewoolie> ka6sox-laptop: I'm waiting on a resistor, of all things.
Jan 12 22:29:34 <ka6sox-laptop> ha!
Jan 12 22:29:37 <beewoolie> One of the resistors on the device failed.
Jan 12 22:29:42 <beewoolie> It broke in two.
Jan 12 22:29:49 <ka6sox-laptop> weird.
Jan 12 22:29:55 <beewoolie> I replaced it with one from the most recent order.
Jan 12 22:29:56 <ka6sox-laptop> thanks for fixing it.
Jan 12 22:30:09 <beewoolie> Luckily I bought a couple extras...erm...194 extras.
Jan 12 22:30:31 <ka6sox-laptop> ha ha!
Jan 12 22:30:34 <ka6sox-laptop> thats funny.
Jan 12 22:30:34 <beewoolie> NP.  It was kinda cool that I figured it out.  I mean, I'm no wizz with the hardware.
Jan 12 22:30:44 <beewoolie> It was only $3 for 200.
Jan 12 22:30:57 <beewoolie> Or, may be $4.  Still, hardly worth buying anything less.
Jan 12 22:31:03 <ka6sox-laptop> ya
Jan 12 22:31:18 <beewoolie> I've been using it to write JTAG code.
Jan 12 22:31:26 <beewoolie> It's interesting how the thing is setup.
Jan 12 22:31:31 <beewoolie> Mostly, it works without too much hassle.
Jan 12 22:31:32 <ka6sox-laptop> are you using linux for fpga stuff?
Jan 12 22:31:45 <beewoolie> There are somethings I don't yet understand.
Jan 12 22:31:46 <ka6sox-laptop> (ie the ISE in Linux?)
Jan 12 22:32:02 <beewoolie> ka6sox-laptop: No, I'm working to replace the functionality of the BDI.
Jan 12 22:32:35 <ka6sox-laptop> w00t! thats what I want too!
Jan 12 22:32:42 <beewoolie> ISE...um...International Securities and Exchange?
Jan 12 22:32:55 <beewoolie> Yeah.  I'm at the stage of reading lots of scan chains.
Jan 12 22:32:59 <ka6sox-laptop> the FPGA developement environment.
Jan 12 22:33:28 <beewoolie> I'm not yet sure how to get access to system registers.  I suspect I need to send ARM instructions over one of the JTAG chanins.
Jan 12 22:33:30 <beewoolie> Ah.
Jan 12 22:33:44 <beewoolie> I don't think I need to learn yet another set of skills...right now.
Jan 12 22:33:58 <beewoolie> I'm OK with getting some high qwality JTAG going.
Jan 12 22:34:03 <ka6sox-laptop> I hope to use the S3e and ep1220's board.
Jan 12 22:35:19 <beewoolie> Interesting note, the fastest we can signal from my laptop is about 260KHz.
Jan 12 22:35:54 <ka6sox-laptop> not very fast.
Jan 12 22:35:57 <beewoolie> With the parallel interface.
Jan 12 22:36:00 <beewoolie> No, not fast.
Jan 12 22:36:03 <beewoolie> That's without delays.
Jan 12 22:36:27 <beewoolie> Also, I now know why the openwince code suxks so badly when we change the frequency.
Jan 12 22:36:42 <ka6sox-laptop> oh? why?
Jan 12 22:36:55 <beewoolie> The usleep () call they use to slow down signaling makes the frequency no greater than 250Hz.
Jan 12 22:37:14 <beewoolie> Sound familiar.  Yep, that's the 10ms limit to the jiffies timer.
Jan 12 22:37:22 <ka6sox-laptop> heh
Jan 12 22:37:36 <beewoolie> In order to make the signaling more adjustable, I'll have to do something fancier.
Jan 12 22:37:49 <beewoolie> Like the bogomips loop.
Jan 12 22:38:32 <beewoolie> Also, openwince is really dumb about detecting cores.
Jan 12 22:38:36 <ka6sox-laptop> I remember playing with this when we were trying to figure out why the Slugs were sooooo slow.
Jan 12 22:38:47 <beewoolie> If there is nothing on the chain, it takes a really long time to detect it.
Jan 12 22:39:14 <beewoolie> Are you referring to our speed-up of the slug?  or of the sluggish upload over jtag?
Jan 12 22:39:19 <beewoolie> no pun meant.
Jan 12 22:39:30 <ka6sox-laptop> the slugs were running at 1/2 speed.
Jan 12 22:39:38 <beewoolie> Right.
Jan 12 22:39:44 <beewoolie> That's was a welcome discovery.
Jan 12 22:40:08 <ka6sox-laptop> okay so if we use the S3e to do the heavy lifting of talking to the chains.
Jan 12 22:40:21 <beewoolie> it turns out that methods for dealing with the IXP core are quite different from methods using on the others.  I'm been working with an ARM9 since I have so many.
Jan 12 22:40:39 <beewoolie> yeah. it would be awesome to get somthing that could signal as fast as 40MHz.
Jan 12 22:41:06 <beewoolie> My code is being designed to construct buffers of signals.  I compose a command list and then process it by a driver.
Jan 12 22:41:16 <beewoolie> I mean, it is designed that way.
Jan 12 22:41:25 <beewoolie> It works pretty well and the code is really simple.
Jan 12 22:41:44 <beewoolie> The next phase will be to get the USB chip to accept a translation of these commands.
Jan 12 22:42:42 <ka6sox-laptop> cool
Jan 12 22:45:04 <ka6sox-laptop> even if we come up with replacement hardware (for the BDI) we still will need a good user interface :)
Jan 12 22:45:38 <ka6sox-laptop> the EP1220 boards are essential to allow it to be universal.
Jan 12 22:48:03 <beewoolie> There is a bothersome ommission on his design.
Jan 12 22:48:06 <beewoolie> We really need sRST.
Jan 12 22:48:11 <beewoolie> nSRST.
Jan 12 22:48:26 <beewoolie> I hope that there is a extra GPIO line we can commandeer.
Jan 12 22:48:37 <ka6sox-laptop> this is the first rev...and I think that I can make that work with the S3e
Jan 12 22:48:47 <ka6sox-laptop> I hope to have mine in Feb.
Jan 12 22:48:52 <beewoolie> Great.  
Jan 12 22:49:02 <beewoolie> You mean, you expect an S3e in Feburary?
Jan 12 22:50:16 <ka6sox-laptop> yes
Jan 12 22:50:28 <ka6sox-laptop> I already have the board from EP1220
Jan 12 22:50:34 <ka6sox-laptop> they are really nice.
Jan 12 22:53:21 <ka6sox-laptop> next time we see him I need to get with him.
Jan 12 22:56:47 <beewoolie> Have you met him before?
Jan 12 22:57:07 <ka6sox-laptop> no...we have just spoken here and via email :)
Jan 12 22:57:29 <beewoolie> I look forward to the day we have our first conference.
Jan 12 22:57:58 <ka6sox-laptop> that would be fun.
Jan 12 22:59:05 <ka6sox-laptop> I think that I'll be able to get a Schematic for the S3e board real soon now :) I need it so that I can start working with the memory controller.
Jan 12 22:59:22 <ka6sox-laptop> they have 32MB of sdram on board.
Jan 12 23:00:57 <beewoolie> That ought to be plenty.
Jan 12 23:01:23 <beewoolie> It would be handy to come up with a protocol so that the S3e could be somewhat autonomous.
Jan 12 23:02:35 <ka6sox-laptop> thats what I want to do.
Jan 12 23:03:04 <ka6sox-laptop> make it so that we can set breakpoints and do other things with the S3e being a little independent.
Jan 12 23:03:12 <beewoolie> One of the things I've toyed with is code to follow the state of the TAP.  That's what the openwince version does.
Jan 12 23:03:18 <beewoolie> However, I've not found it necessary.
Jan 12 23:03:51 <beewoolie> Actually, I was thinking that the S3e would only know about JTAG.  The details of how debugging is done with be handled by the host.
Jan 12 23:04:05 <beewoolie> If we want to put the debug info into the device, we'll have to specialize for every target.
Jan 12 23:04:22 <beewoolie> On the other hand, the JTAG scanning is standardized.
Jan 12 23:04:26 <ka6sox-laptop> yep
Jan 12 23:04:28 <beewoolie> All we need to do is feed it commands.
Jan 12 23:04:35 <ka6sox-laptop> okay lets start simple :)
Jan 12 23:04:42 <beewoolie> The handy thing would be an ability to download fragments and call them out.
Jan 12 23:04:50 <ka6sox-laptop> 1st thing is FAST programming
Jan 12 23:04:55 <beewoolie> There is a 'reset TAP' fragment that is really obvious.
Jan 12 23:05:07 <beewoolie> Others require scanning through several chains to get a result.
Jan 12 23:05:10 <beewoolie> Macro-like.
Jan 12 23:05:14 <beewoolie> Right.
Jan 12 23:05:19 <ka6sox-laptop> then 2ond we can start scanning the chain for things :
Jan 12 23:05:24 <beewoolie> OK, here's my idea for that.
Jan 12 23:05:46 <beewoolie> There is a 32 bit port available on the ARM cores.
Jan 12 23:05:51 <beewoolie> We can send data back and forth.
Jan 12 23:06:08 <beewoolie> Well, we could write a program for the ARM that looks at this port and writes the data to flash.
Jan 12 23:06:32 <beewoolie> It adds an extra step, but it allows us to program as fast as the device can go.
Jan 12 23:07:01 <beewoolie> It also means that we don't have to write anything more that the data to program into the device.
Jan 12 23:07:32 <beewoolie> Plus, the program running on the core can buffer data, overlapping IO with the JTAG device with programming to flash.
Jan 12 23:07:52 <beewoolie> If we make it possible to perform buffered writes, we should be able to go blazing fast.
Jan 12 23:08:15 <beewoolie> The really slick method for doing this would be to put the code into cache.
Jan 12 23:08:24 <beewoolie> Then, we wouldn't even need RAM.
Jan 12 23:11:44 <beewoolie> Nice chatting.
Jan 13 01:18:14 <ka6sox-laptop> [23:07] 	<ka6sox-laptop>	they mis labeled the package and therefore they sent it the "wong way"
Jan 13 01:18:14 <ka6sox-laptop> [23:07] 	<ka6sox-laptop>	er wrong way!
Jan 13 01:18:14 <ka6sox-laptop> [23:08] 	<ka6sox-laptop>	we need to call 800-622-4205 (between 7am and 6pm CST) to complain and ask for the shipping charges back.
Jan 13 01:18:14 <ka6sox-laptop> [23:08] 	<ka6sox-laptop>	they *will* deliver the package tommorrow.
Jan 13 01:18:16 <ka6sox-laptop> [23:08] 	<ka6sox-laptop>	it has been re-routed to Goleta and they did verify my address and phone number.
Jan 13 01:18:18 <ka6sox-laptop> [23:28] 	<--	dyoung-away has quit (Read error: 110 (Connection timed out))
Jan 13 01:18:21 <ka6sox-laptop> eeek
**** ENDING LOGGING AT Fri Jan 13 02:59:57 2006